摘要
针对中文人称名词短语单复数属性信息不明确,对消解贡献低的问题,利用改进的最大熵模型提出了人称代词消解新的模型。该模型在原有特征的基础上增加了人称名词短语单复数识别的Head特征、Qun特征和Len特征。在真实文本环境下与不使用单复数属性信息方法进行了对比实验,结果表明该方法的F值与不使用单复数属性信息方法的F相比有一定的提高。
To solve the problem of whether Chinese personal pronouns noun phrase is singular or pluralism is not obvious, which contributes less to anaphora resolution, the paper presents a new model for personal pronouns anaphora resolution based on improved maximum entropy model. The model mainly adds Head character, Qun character and Len character to original characteristics that can recognize the personal pronouns noun phrase is singular or plural. The contrast experiment shows that the algorithm presented in the paper has better F value.
出处
《江南大学学报（自然科学版）》
CAS
2009年第6期666-669,共4页
Joural of Jiangnan University (Natural Science Edition)
基金
江苏省研究生创新计划项目(CX098-203Z)
关键词
人称代词消解
人称名词短语单复数
最大熵
指代消解
personal pronouns resolution, person singular and plural noun phrase, maximum entropy, anaphora resolution
